Vibration Data Processing Workflow
How to process accelerometer and vibration test data — FFT, PSD, frequency response, windowing and sampling considerations.
Engineering Task
Processing vibration test data to extract frequency content, PSD or frequency response functions for correlation with FE models.
Workflow
- Verify sampling rate — must be at least 2.5× the highest frequency of interest (Nyquist)
- Apply windowing (Hanning) to reduce spectral leakage in FFT
- Compute FFT or PSD using Welch averaging for statistical reliability
- For FRF: compute output/input cross-spectrum and input auto-spectrum
- Check coherence — low coherence indicates noise, leakage or nonlinearity
Key Considerations
- Sampling rate: at least 2.5× (practically 5×) the highest frequency of interest
- Windowing: Hanning for general use; flattop for amplitude accuracy
- Averaging: use at least 8–16 averages for stable PSD or FRF estimates
- Units: verify g²/Hz for PSD, g/N for FRF — unit errors are common
- Sensor calibration: apply sensitivity (mV/g) from the calibration certificate